|Publication number||US8149911 B1|
|Application number||US 11/675,715|
|Publication date||Apr 3, 2012|
|Filing date||Feb 16, 2007|
|Priority date||Feb 16, 2007|
|Publication number||11675715, 675715, US 8149911 B1, US 8149911B1, US-B1-8149911, US8149911 B1, US8149911B1|
|Inventors||Josť R. Alvarez, Guy Cote, Udit Budhia|
|Original Assignee||Maxim Integrated Products, Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (22), Non-Patent Citations (9), Referenced by (7), Classifications (6), Legal Events (6)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The present invention relates to video processing generally and, more particularly, to a method and/or apparatus for implementing a multiple pass digital image stabilization system.
As camcorders and other video recording devices (i.e., digital still cameras, mobile phones, etc.) continue to shrink in size, and as zoom ratios continue to increase, it becomes increasingly difficult for users to steadily hold a camera to produce stable video.
Camera ergonomics may not allow holding the device in a stable and comfortable position and thus promote unstable holding of the device. Also, because of the highly mobile nature of these devices, people are increasingly capturing video in less than ideal situations (i.e., outdoor activities, sporting events, etc.) as opposed to contrived in-door events. Therefore, there is less opportunity for properly supporting the camera during recording.
Furthermore, as optics continue to improve, magnification capabilities are often incorporated in such devices. High magnification factors (i.e., zooming) contribute to the unstable appearance of video since such zooming amplifies every small movement of the hand of the user.
Camera jitter (i.e., mechanical instability) introduces extraneous motion during video capture. The extraneous motion is not related to the actual motion of objects in the picture. Therefore, the motion appears as random picture movements that produce disturbing visual effects. The motion can be difficult to encode at low bit rates. The end result is video material that is hardly usable from both practical and aesthetic perspectives.
Camcorder manufacturers have implemented various ways of implementing image stabilization. One way is to use mechanical correction, including piezo-electric physical displacement, optical system fluid coupling/dampening and other mechanical dampening devices. Another way of solving the problem is by electronic correction (i.e., digital signal processing) using external sensors.
Current digital image stabilization solutions are limited by the type of processors used in typical cameras. These processors are more geared toward Image/Sensor Processing and therefore do not have easy access to the sophisticated motion estimation statistics commonly available in hybrid entropy video encoder/decoders (Codecs). Furthermore, in cases when a digital stabilization is used in the context of a video Codec, a large number of motion vectors are used in a single pass without a flexible selection of areas of motion and in a non-hierarchical motion estimation architecture.
It would be desirable to remove extraneous motion from an input video signal to produce a stabilized sequence of pictures that is more visually pleasing and/or more easily compressed.
The present invention concerns an apparatus including a first circuit and a second circuit. The first circuit may be configured to generate (i) a first series of sequential frames, (ii) a plurality of local motion vectors for each of said frames, (iii) one or more global motion vectors for each of said frames, (iv) a second series of stabilized sequential frames, (v) a plurality of rough motion vectors and (vi) a digital bitstream in response to (i) a video input signal. The second circuit may be configured to store (i) the first series of sequential frames, (ii) the plurality of local motion vectors, (iii) the one or more global motion vectors, (iv) the second series of stabilized sequential frames and (v) the plurality of rough motion vectors.
The objects, features and advantages of the present invention include providing a method and/or apparatus for implementing a digital image stabilization system that may (i) use available pre-processing structures (e.g., cropping, polyphase scaling, statistics gathering, feature classification regions, etc.), (ii) allow flexibility in using external overscan sensors such that the order of cropping and scaling during motion compensation may be reversed (or to allow scaling to be bypassed completely), (iii) implement a hierarchical motion estimation architecture that allows localized sets of motion vectors to be flexibly defined at any spatial location in a picture, (iv) implement pre-motion estimation that may be performed in the subsampled picture domain in order to increase motion detection range, (v) implement full-pel accurate motion vectors, (vi) achieve sub-pel global compensation through a scaling process, (vii) allow multiple pass analysis and detection of the image sequence to improve quality in an analogous manner as dual-pass rate control, (viii) reduce the data set of local motion vectors to simplify global motion vector computations, (ix) allow picture sequence adaptivity by analyzing statistical data gathered on both processing paths, (x) provide adaptivity that may be achieved for local motion as well as global motion by time-series processing of resulting stabilization data, (xi) allow encoding statistics to be used in determination of best quality in multiple encoding passes (e.g., does not preclude the use of multiple fast encodings and multiple stabilization enhancements) and/or (xii) provide stabilization by implementing multiple (e.g., recursive) processing passes.
These and other objects, features and advantages of the present invention will be apparent from the following detailed description and the appended claims and drawings in which:
The present invention relates to providing a system that may be used to stabilize captured images in order to improve visual quality and/or the amount of compression. The present invention may use one or more digital image stabilization (DIS) techniques. In one example implementation, digital signal processing (DSP) may be used to estimate and compensate for random jitter introduced by the movement of a camera (or other capture device) during operation.
The block 102 may have an input 110 that may receive a signal (e.g., P) and an output 112 that may present a signal (e.g., BITSTREAM). The signal P generally represents an unencoded video input signal. In one example, the signal P may be received from an imaging sensor, or other capture device. The signal BITSTREAM generally represents an encoded digital bitstream. The signal BITSTREAM may be implemented, in one example, as a compressed bitstream. The signal BITSTREAM may be compliant with one or more standard or proprietary encoding/compression specifications.
In one example, the block 102 may have an input 114 that may receive a signal (e.g., C), an output 116 that may present a signal (e.g., STATS), an output 118 that may present the signal P, an input 120 that may receive a signal (e.g., DIS_SW), an input 122 that may receive a signal (e.g., ME_SW), an output 124 that may present a signal (e.g., RMV), an output 126 that may present a signal (e.g., LMV) and an input 128 that may receive a signal (e.g., SBS). In one example, the memory 104 may have an output 130 that may present the signal C, an input 132 that may receive the signal STATS/P, an input 134 that may receive the signal P, an output 136 that may present the signal DIS_SW, an output 138 that may present the signal ME_SW, an input 140 that may receive the signal RMV, an input 142 that may receive the signal LMV and an output 144 that may present the signal SBS. The signal C may comprise one or more cropped images (or pictures). The signal STATS/P may comprise stabilized picture and statistics information. The signal P may comprise unstable (e.g., jittery) input video information. The video information in the signal P may be full resolution (e.g., capture resolution). The signal DIS_SW may comprise search window information (e.g., location, search ranges, number of search areas and any other parameters specified by the digital image stabilization technique implemented by the circuit 102). The signal ME_SW may comprise information that may be used in perform a motion estimation process compliant with an encoding process implemented by the circuit 102. The signal RMV may comprise rough motion vector information. The signal LMV may comprise local motion vector information. The signal SBS may comprise stabilized picture and statistics information that may be used by the encoding process implemented by the circuit 102.
The inputs, outputs and signals shown coupling the block 102 and the block 104 generally represent logical inputs, logical outputs and logical data flows. The logical data flows are generally illustrated as signals communicated between respective the inputs and outputs for clarity. As would be apparent to those skilled in the relevant art(s), the inputs, outputs, and signals illustrated in
The circuit 150 may receive a signal (e.g., GMV), the signal P and the signal C. The circuit 150 may generate a signal (e.g., STATS), the signal P and the signal STATS/P in response to the signals GMV, P and C. The circuit 152 may receive a signal (e.g., STATS2) and the signals STATS, STATS/P, RMV and LMV. The circuit 152 may generate a signal (e.g. CTRL) and the signal GMV in response to the signals STATS, STATS2, STATS/P, RMV and LMV. The circuit 154 may receive the signals CTRL, DIS_SW and ME_SW. The circuit 154 may generate the signals RMV and LMV in response to the signals CTRL, DIS_SW and ME_SW. The circuit 156 may receive the signal SBS. The circuit 156 may generate the signal BITSTREAM and the signal STATS2 in response to the signal SBS.
The system 100 may process the input video data signal P in two stages. The unstable (jittery) input video signal P may be passed through the video preprocessor circuit 150 at full resolution to capture the signal P as a sequence of video pictures. The video preprocessor circuit 150 may generate statistics for the unstable input video signal P that may be communicated (e.g., via the signal STATS) to the controller 152 for analysis. The analysis may include, for example, scene detection and sudden event detection. Pictures containing unstable portions of the input video signal P may be stored in the memory 104 (e.g., via the signal P) for further processing.
The premotion estimator circuit 154 may receive the location, search ranges, number of areas and other parameters from the memory 104 (e.g., via the signal DIS_SW) and the controller 152 (e.g., via the signal CTRL) based on indications (or instructions) from the firmware (or software) executed by the controller 152. The premotion estimator circuit 154 may use the location, search ranges, number of areas and other parameters indicated by the firmware executed by the controller 152 to compute and transmit raw local motion vectors. The raw local motion vectors may be computed for a specific block or set of blocks in the picture being processed. The raw local motion vectors may be presented to the memory 104 and the controller 152 via the signal LMV. Further processing of the raw local motion vectors may produce the GMV that eventually is used to compensate (stabilize) the picture.
The controller 152 analyses the signal STATS and the local motion vectors (LMVs) for a respective picture to produce a global motion vector (GMV) and other control information for motion compensation (stabilization) of the respective picture. The video preprocessor 150 receives the global motion vector and other control information for motion compensation (stabilization) of the respective picture via the signal GMV and retrieves one or more cropped pictures from a displaced location in the memory 104. The displaced location for retrieving the cropped picture(s) is generally indicated by the global motion vector(s) received from the controller 152. In one example, the video preprocessor circuit 150 may perform scaling with sub-phase accuracy (e.g., using a multiple phase scaler) to produce a sub-pel displacement (stabilization). The video preprocessor circuit 150 writes the stabilized picture(s) and statistics to the memory 104 and the controller 152 (e.g., via the signal STATS/P).
The premotion estimator circuit 154 retrieves information from the memory 104 (e.g., via the signal ME_SW) for performing a pre-motion estimation process. The pre-motion estimation process may be performed in the normal hierarchical motion estimation process which is normally part of the encoder process (e.g., H.264, etc.) implemented by the system 100. The premotion estimator circuit 154 writes search area rough motion vectors (RMV) to the memory 104 and the controller 152 (e.g., via the signal RMV). The rough motion vectors may be used by the encoder 156 for further motion estimation refinement.
The encoder 156 uses the stabilized picture and statistics information retrieved from the memory 104 (e.g., via the signal SBS) to code pictures with the best quality. The encoder 156 produces encoded bitrate and compression statistics that are communicated to the controller 152 (e.g., via the signal STATS2). The encoded bitrate and compression statistics may be used for further refinements to the digital image stabilization process.
The system 100 may provide digital image stabilization (DIS) using digital information extracted from the input video signal P. The system 100 may perform sub-pel accurate DIS through the video preprocessor circuit 150, the premotion estimator circuit 152 and a method (or process) running on the controller 152. The controller 152 may be implemented as a programmable processor. The system 100 may generate a global motion vector (GMV) for each picture obtained through the stabilization process. In one example, the digital stabilization process may be implemented in software or firmware. For example, the digital stabilization process may be implemented and/or controlled using computer executable instructions stored in a computer readable medium.
The video preprocessor circuit 150 and the premotion estimator circuit 154 may be used during encoding operations. The video preprocessor circuit 150 and the premotion estimator circuit 154 may also be used in parallel under firmware control to compute the global motion vectors 210 a-210 n and/or to perform global image displacement for improved coding.
The actual global displacement indicated by the global motion vector GMV may use sub-pel accuracy. In order to perform sub-pel accurate displacement two mechanisms may be implemented. The first one comprises reading an offset location from memory. For example, if the original picture is stored at location x,y in the memory 104, a GMV (global motion vector) may be generated that indicates the image should be read from location (x+n, y+m), where the n,m value is the two-dimensional displacement. If the actual displacement computed turns out to be a fractional number (e.g., a non-integer pixel displacement) the non-integer part of the displacement may be computed by interpolation using a polyphase filter.
Each local motion vector (LMV) 206 a-206 n may be the result of the plurality of motion vectors 204 a-204 n derived from adjacent blocks (e.g., macroblocks) in the respective search windows 202 a-202 n. In one example, a single LMV 206 may be derived for each local cluster of motion vectors 204 a-204 n. The single LMV 206 may be used for further processing. In a preferred embodiment, a recursive method may be used to derive the local and global vectors (described below in connection with
In one example, the motion vectors may be processed in such a way as to remove DC components. The AC (varying) components of the global motion vectors may be used to determine the motion shift used to compensate for shaking of the pictures. The system 100 may be used to perform temporal processing, smoothing and prediction through median filtering, clustering, and/or averaging. In addition, a kalman filter predictor may be incorporated as part of the temporal filtering to further stabilize the results.
For example, N processed GMVs 210 a-210 n may be gathered (e.g., one for each picture 200 a-200 n). A temporal filtering operation may be performed on the N consecutive GMVs. In one example, the temporal filtering may comprise a temporal averaging of the samples. The value resulting from the temporal averaging constitutes a DC component for the set of N samples. The DC value is removed from the current GMV in proportion using a simple predictor (e.g., convex operation as indicated below, or a kalman filter). In a preferred embodiment a running sum (e.g., accumulator Acc(t)) of the motion vectors may be maintained as each vector is obtained. Maintaining the running sum is equivalent to performing a translational shift with respect to the first frame in the sequence of N samples.
In one example, the convex operation may be implemented according to the following Equation 1:
DC(t)=((1−alpha)*DC(t−1))+(alpha*Acc(t)) Eq. 1.
A final displacement for time ‘t’ may be expressed by the following Equation 2:
GMV_shift(t)=GMV(t)−DC(t), EQ. 2
where GMV_shift(t) is the actual value used to compensate the picture for stabilization. The goal is that the final shift used to globally compensate for ‘shaking’ be of zero mean over a period of time.
The process 250 may comprise a state (circuit) 252, a state (circuit) 254, a state (circuit) 256, a state (circuit) 258, a state (circuit) 260 and a state (circuit) 262. The state 252 may be implemented as a motion vector scanning process. The state 252 may arrange multiple motion vectors by performing a serialization scan (e.g., zig-zag, reversible, spiral, etc.). The state 254 may be implemented as a median filter state. In one example, the state 254 may be implemented as a 5-tap median filter. However, other numbers of taps (e.g., 3, etc.) may be implemented accordingly to meet the design criteria of a particular implementation. The state 256 may be implemented as a center predictor state. The state 258 may be implemented as a multiplexing state. The state 260 may be implemented as a post process filter state. The state 260 generally converts a cluster of motion vectors into a single motion vector. In one example, the conversion may include a linear combination of the motion vectors. In a preferred embodiment, the linear combination may be implemented as an averaging operation of the motion vectors. The state 262 may be implemented as a recursive control state.
The state 252 may receive multiple motion vectors of full-pel or greater (e.g., sub-pel such as half-pel, quarter-pel, eighth-pel, etc.) accuracy. The state 252 may present the motion vectors to an input of the state 254, a first input of the state 256 and a first input of the state 258 in an order that reduces entropy between the multiple motion vectors. The state 254 performs a median filtering operation on the ordered MVs and presents the result to a second input of the state 258. In a preferred embodiment the state 254 uses a 5-tap median filter.
The state 256 receives a threshold value (e.g., THRESHOLD) at a second input and generates a control signal in response to the ordered MVs and the threshold value. The state 256 presents the control signal to a control input of the state 258. The state 258 selects either the output from the state 254 or the ordered MVs from the state 252 for presentation to an input of the state 260 in response to the control signal received from the state 256.
The state 256 performs a post process filtering operation on the output of the state 258. In one example, the post processing operation comprises performing an averaging filter on the MVs received from the state 258. When the multiple motion vectors received by the state 252 comprise motion vectors for a cluster, the state 260 presents a single average MV for the entire cluster (e.g., a local motion vector (LMV) 206). The single average MV generated by the state 260 is presented to an input of the state 262.
When all of the LMVs for each chosen location (e.g., search window) in the picture are obtained, the multiple LMVs may be presented to the input of the state 252 and the process 250 may be performed on the LMVs to generate a global motion vector (GMV) 210 (e.g., a recursive step). The LMVs and GMVs generated by the process 250 may be generated with sub-pel accuracy, even when the input motion vectors presented to the state 252 have only full-pel accuracy. In a preferred embodiment, only full-pel accurate motion vectors are used for generation of LMVs and GMVs in order to reduce computational demands. In general, both local and rough motion vectors may be sub-pel or full-pel accurate, depending upon the design criteria of a particular implementation. The best choice for quality is sub-pel accuracy, because sub-pel accuracy means the reach of the motion vector is between pixels, and therefore more accurate. The best choice for power utilization/processing time is full-pel accuracy, because there are fewer samples to process.
As used here-in ‘full’ generally means ‘not fractional’. For example, in a design with full-pel accurate motion vectors, the motion vectors may be referred to as mv0, mv1, mv2, mv3. In a design with sub-pel accurate motion vectors, for example, quarter-pel, instead of having four vectors as listed above, thirteen vectors would be implemented (e.g., mv0.0, m0.25, mv0.50, mv0.75, mv1.0, mv1.25, mv1.50, . . . , mv2.75, mv3.0). Although the input motion vectors to the stabilization process 250, (e.g., the local MVs and rough MVs) may have only full-pel accuracy, the output of the process 250 (e.g. the LMVs or GMVs) which are computed from local MVs and rough MVs (e.g., as a linear combination of the local MVs and rough MVs), generally have sub-pel accuracy, regardless of the accuracy of the input motion vectors.
The global motion vectors 210 a-210 n may be used to modify (adjust) the particular encoding process (e.g., H.264 or other) implemented. While the method 300 provides a modification to the encoding process, the signal BITSTREAM is generally generated as a compliant bitstream that may be decoded by any compliant decoder (e.g., an H.264 or other decoder).
If the local cluster of blocks chosen to generate the motion vectors is positioned in a flat area of the picture (e.g., an area where there is very little detail) and there is little real motion in the area, the calculation of local motion vectors will produce motion vectors that are not reliable. For example, the difference between the block under processing and the reference block may be very small when an area where there is very little detail and/or little real motion and therefore any block will produce an acceptably low error, which is not indicative of motion.
To ensure reliable motion vectors are generated, image statistics may be obtained from the video preprocessor block 150. The image statistics may include spatial low and high frequency as well as edge information. Given a bandwidth threshold of 0.5 Nyquist, a block that has frequency information below the threshold may be classified as ‘low frequency’ and a block that has information above the threshold may be classified as ‘high frequency’. The average value of all the pixels in the block below and above the bandwidth threshold represents the amount of such feature in the block. Similarly, the output of an edge detector performed on the pixels in the block may be averaged over all the pixels in the block, and the result used as an indication of edge energy in the block.
In one example, a location for the cluster of blocks may be chosen that has more than 10% high frequency, less than 50% of low frequency and strong edges (e.g., more than 5%). If an area of the picture meets the above criteria, the area may be chosen as a possible candidate for clustering. Once all blocks in the picture are examined, the actual areas may be chosen. The decision may be based on system limitations, but experience has shown that nine clusters are sufficient for good results.
In general, an encoder may be made more efficient by receiving a stabilized sequence of pictures. The increased efficiency may be translated into lower power (e.g., fewer computations performed) since the motion estimation range may be lowered for a stabilized picture. The increased efficiency may also be translated into smaller bitstreams since a stabilized sequence may be easier to encode. For example, the stabilized sequence may produce smaller compressed sizes compared to those produced by unstable (shaking or jittery) sequences, while preserving a high level of visual quality. In general, more bits would be used to maintain a high level of visual quality in an unstable sequence than would be used in a sequence stabilized in accordance with the present invention.
Traditional camcorder and DSC companies are likely to include DIS as part of their sensor and image processing pipeline, which are normally not adept at motion estimation processing. Conventional solutions are replete with shortcuts and compromises for the critical global motion estimation process. The system 100 takes advantage of the sophisticated mechanisms for motion estimation in hybrid entropy codecs, and statistical data gathered at various stages of the process for quality refinements. The system 100 may also decouple the global estimation (e.g., rough search) from the refined motion estimation thereby allowing flexible parallel reuse of each module.
The present invention may be used to minimize the cost of image stabilization in an integrated image stabilization processor/video encoder solution by re-using motion estimation designed for video encoding to achieve digital image stabilization. The present invention may be used to provide an effective image stabilization system as part of the normal video coding pipeline that may reduce costs by reusing and repurposing parts of the processing chain.
The present invention may be used in products that may straddle the boundary of image stabilization and video coding. The present invention, by virtue of system on a chip (SOC) integration, may be used in products that integrate image stabilization. The present invention may be used to allow us to a flexible and/or scalable solution for current and future products.
When looking for the best match for a block in the current picture, the proper selection of the starting search position in a reference picture is important to the success of a practical motion estimation process. In order to make a proper selection, a premotion estimator may be implemented to make a rough estimate of the starting position from which to perform a refined motion estimation. The rough estimation may be done in a multiple-pixel domain to obtain an estimate of where to start a refinement. The refined estimation may be made with sub-pel match accuracy and therefore require large amounts of computation. Although it is possible to use only the refined motion estimation, it is more efficient to perform a hierarchical search with a premotion estimator first.
The present invention may be implemented to essentially repurpose the resulting vectors from the premotion estimator to perform image stabilization. When the premotion estimator is configured so that the resulting rough motion vectors favor the co-located block position and are performed so that the cost from distortion is more important than the cost of coding the motion vector, the resulting vector may be used as a facsimile of true motion.
A premotion estimator module for performing rough estimation may be programmed according to the teaching of the present invention to produce localized motion vectors. A control layer may be implemented (e.g., in software, firmware, etc.) to process a plurality of localized premotion estimation vectors to produce a single global motion vector (GMV) for the picture. A plurality of GMVs for a video sequence of pictures may be further processed in accordance with the present invention to produce a stabilization displacement for every picture in the video sequence.
While the invention has been particularly shown and described with reference to the preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5999651 *||Jun 6, 1997||Dec 7, 1999||Matsushita Electric Industrial Co., Ltd.||Apparatus and method for tracking deformable objects|
|US6084912 *||Jun 27, 1997||Jul 4, 2000||Sarnoff Corporation||Very low bit rate video coding/decoding method and apparatus|
|US6628711 *||Jul 2, 1999||Sep 30, 2003||Motorola, Inc.||Method and apparatus for compensating for jitter in a digital video image|
|US6809758 *||Dec 29, 1999||Oct 26, 2004||Eastman Kodak Company||Automated stabilization method for digital image sequences|
|US6968009 *||Nov 8, 2000||Nov 22, 2005||Stmicroelectronics, Inc.||System and method of finding motion vectors in MPEG-2 video using motion estimation algorithm which employs scaled frames|
|US7408986 *||Jun 13, 2003||Aug 5, 2008||Microsoft Corporation||Increasing motion smoothness using frame interpolation with motion analysis|
|US20020118761 *||Jun 25, 2001||Aug 29, 2002||Samsung Electronics Co., Ltd.||Decoder having digital image stabilization function and digital image stabilization method|
|US20040120197 *||Jan 10, 2003||Jun 24, 2004||Tetsujiro Kondo||Memory cell circuit, memory device, motion vector detection device, and motion compensation prediction coding device|
|US20060017814 *||Jul 21, 2004||Jan 26, 2006||Victor Pinto||Processing of video data to compensate for unintended camera motion between acquired image frames|
|US20060023790 *||Dec 29, 2004||Feb 2, 2006||Industrial Technology Research Institute||Method for processing motion information|
|US20060159311 *||Nov 17, 2005||Jul 20, 2006||Mitsubishi Denki Kabushiki Kaisha||Dominant motion analysis|
|US20060188021 *||Feb 24, 2006||Aug 24, 2006||Sanyo Electric Co., Ltd.||Motion vector detecting device|
|US20060206292 *||Nov 25, 2003||Sep 14, 2006||Walid Ali||Method and apparatus to utilize the probability vectors in the binary representation of video systems for faster convergence with minimal computation requirements|
|US20060228049 *||Feb 14, 2006||Oct 12, 2006||Stmicroelectronics S.A.||Method for capturing images comprising a measurement of local motions|
|US20060257042 *||May 13, 2005||Nov 16, 2006||Microsoft Corporation||Video enhancement|
|US20060290821 *||Aug 30, 2006||Dec 28, 2006||Microsoft Corporation||Video processing system and method for automatic enhancement of digital video|
|US20070076982 *||Sep 30, 2005||Apr 5, 2007||Petrescu Doina I||System and method for video stabilization|
|US20070092111 *||Aug 31, 2004||Apr 26, 2007||Wittebrood Rimmert B||Motion vector field re-timing|
|US20070154066 *||May 22, 2006||Jul 5, 2007||Industrial Technology Research Institute||Object tracking systems and methods|
|US20080004073 *||Jun 30, 2006||Jan 3, 2008||Motorola, Inc.||Methods and devices for video correction of still camera motion|
|US20080030586 *||Aug 7, 2006||Feb 7, 2008||Rene Helbing||Optical motion sensing|
|JPH11266459A *||Title not available|
|1||A.J. Crawford, et al., "Gradient Based Dominant Motion Estimation With Integral Projections For Real Time Video Stabilisation", IEEE, 2004, pp. 3371-3374.|
|2||Dietmar Wueller, "Evaluating Digital Cameras", SPIE-IS&T/vol. 6069 60690K-1, 2006, 15 pages.|
|3||Filippo Vella et al., "Digital Image Stabilization by Adaptive Block Motion Vectors Filtering", IEEE Transactions on Consumer Electronics, vol. 48, No. 3, Aug. 2002, pp. 796-801.|
|4||Haruhisa Okuda et al., "Optimum Motion Estimation Algorithm for Fast and Robust Digital Image Stabilization", IEEE Transactions on Consumer Electronics, vol. 52, No. 1, Feb. 2006, pp. 276-280.|
|5||Ikuko Tsubaki et al., "An Adaptive Video Stabilization Method For Reducing Visually Induced Motion Sickness", IEEE, 2005, pp. III-497 through III-500.|
|6||Jie Shao et al., "Simultaneous Background And Foreground Modeling For Tracking In Surveillance Video", IEEE, 2004, pp. 1053-1056.|
|7||Kenya Uomori et al., "Automatic Image Stabilizing System by Full-Digital Signal Processing", IEEE Transactions on Consumer Electronics, vol. 36, No. 3, Aug. 1990, pp. 510-519.|
|8||Marius Tico et al., "Constraint Motion Filtering for Video Stabilization", IEEE, 2005, pp. III-569 through III-572.|
|9||Yu-Chun Peng et al., "Digital Image Stabilization and Its Integration With Video Encoder", IEEE, 2004, pp. 544-549.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US8923400 *||Nov 14, 2007||Dec 30, 2014||Geo Semiconductor Inc||Method and/or apparatus for multiple pass digital image stabilization|
|US9224212 *||Sep 6, 2013||Dec 29, 2015||Canon Kabushiki Kaisha||Image processing apparatus and image processing method|
|US9361545 *||Jun 4, 2014||Jun 7, 2016||Qualcomm Incorporated||Methods and apparatus for estimating angular movement with a single two dimensional device|
|US20110051003 *||Aug 27, 2008||Mar 3, 2011||Powerlayer Microsystems Holding Inc.||Video image motion processing method introducing global feature classification and implementation device thereof|
|US20140072180 *||Sep 6, 2013||Mar 13, 2014||Canon Kabushiki Kaisha||Image processing apparatus and image processing method|
|US20150016683 *||Mar 18, 2013||Jan 15, 2015||Sony Corporation||Image processing device, image processing method, and program|
|US20150139537 *||Jun 4, 2014||May 21, 2015||Qualcomm Incorporated||Methods and apparatus for estimating angular movement with a single two dimensional device|
|U.S. Classification||375/240.12, 375/240.16|
|Cooperative Classification||G06T5/003, G06T7/20, G06T2207/20201|
|Feb 16, 2007||AS||Assignment|
Owner name: MOBILYGEN CORPORATION, CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:ALVEREZ, JOSE R;BUTLER, SIMON;BUDHIA, UDIT;REEL/FRAME:018896/0811
Effective date: 20070215
|Apr 9, 2007||AS||Assignment|
Owner name: MOBILYGEN CORPORATION, CALIFORNIA
Free format text: CORRECTIVE ASSIGNMENT TO CORRECT THE INVENTOR S NAME, PREVIOUSLY RECORDED ON REEL 018896 FRAME 0811;ASSIGNORS:ALVAREZ, JOSE R.;COTE, GUY;BUDHIA, UDIT;REEL/FRAME:019167/0023
Effective date: 20070215
|Sep 25, 2009||AS||Assignment|
Owner name: MAXIM INTEGRATED PRODUCTS, CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MOBILYGEN CORPORATION;REEL/FRAME:023283/0101
Effective date: 20090922
|Jan 23, 2013||AS||Assignment|
Owner name: GEO SEMICONDUCTOR INC., CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MAXIM INTEGRATED PRODUCTS, INC.;REEL/FRAME:029677/0261
Effective date: 20130118
|Oct 23, 2013||AS||Assignment|
Owner name: BISHOPSGATE HOLDINGS CORPORATION, BERMUDA
Free format text: SECURITY AGREEMENT;ASSIGNOR:GEO SEMICONDUCTOR INC;REEL/FRAME:031479/0486
Effective date: 20130809
|Sep 16, 2015||FPAY||Fee payment|
Year of fee payment: 4